„the relationship between immigration and unemployment...

MAGISTERARBEIT

Titel der Magisterarbeit

„THE RELATIONSHIP BETWEEN IMMIGRATION AND UNEMPLOYMENT: THE CASE OF AUSTRIA”

Verfasst von

CIHAN YAYLALI, Bakk.rer.soc.oec

angestrebter akademischer Grad

Magister der Sozial- und Wirtschaftswissenschaften

(Mag.rer.soc.oec.)

Wien, 2014

Studienkennzahl It. Studienblatt: A 066 913

Studienrichtung lt. Studienblatt: Magisterstudium Volkswirtschaftslehre

Betreuer: Univ.-Prof. Dr. Robert M. Kunst

2

CONTENTS

I. LIST OF FIGURES ............................................................................................... 3

II. LIST OF TABLES ................................................................................................. 4

1. INTRODUCTION .................................................................................................. 6

2. THEORETICAL BACKGROUND AND LITERATURE REVIEW ........................... 8

3. IMMIGRATION TO AUSTRIA SINCE THE 1960s .............................................. 15

4. DATA SOURCE AND DESCRIPTIVE STATISTICS .......................................... 26

4.1 ECONOMETRIC MODEL ........................................................................ 27

4.2 UNIT ROOT TESTS ................................................................................ 28

4.3 LONG-RUN RELATIONSHIPS AND COINTEGRATED VARIABLES ..... 30

4.4 SHORT-RUN RELATIONSHIPS .............................................................. 34

4.5 IMPULSE RESPONSE FUNCTIONS ...................................................... 36

5. CONCLUSIONS ................................................................................................. 37

6. REFERENCES ................................................................................................... 39

7. APPENDIX A ...................................................................................................... 43

8. APPENDIX B ...................................................................................................... 46

9. CV ...................................................................................................................... 50

3

I. LIST OF FIGURES

Figure 1: Short-Run Impact of Immigration when Immigrants and Natives are

Perfect Substitutes ............................................................................................ 10

Figure 2: Short-Run Impact of Immigration when Immigrants and Natives are

Complements .................................................................................................... 11

Figure 3: Foreign Population and Foreign Labor Force in Austria (1962-2012) 17

Figure 4: Foreign Share in the Total Population in Austria (1961-2012) ........... 18

Figure 5: Net Migration Balance Rate in Austria (1975-2012) ........................... 19

Figure 6: Naturalized Persons in Austria (1960-2013) ...................................... 21

Figure 7: Unemployment Rates of Natives and Immigrants in Austria (1975-

2012) ................................................................................................................. 22

Figure 8: Net Migration Balance Rate and Unemployment Rate in Austria (1975-

2012) ................................................................................................................. 23

Figure 9: Income by Citizenship in Austria (2003-2012) ................................... 24

Figure 10: Foreign Share by Industry in Austria (2008-2013) ........................... 25

Figure 11: Impulse Response of Unemployment Rate on Migration Rate ........ 36

4

II. LIST OF TABLES Table 1: Descriptive Statistics ........................................................................... 27

Table 2: Stationarity Test in Levels (ADF and PP) ............................................ 29

Table 3: Stationarity Test in First-Differencing (ADF and PP) ........................... 30

Table 4: Johansen’s Trace Test and Maximum Eigenvalue Results ................. 32

Table 5: Results of Adjustment Coefficients from Normalized Co-Integrating

Vectors: ............................................................................................................. 35

5

The Relationship Between Immigration and Unemployment: The Case of Austria.

Abstract

This paper examines empirically the relationship between immigration and the Austrian labor market variables, such as unemployment, real wage and per capita GDP by using the Johansen co-integration procedure, a vector error correction model (VECM), and Impulse Response Functions (IRFs), respectively. The estimation results reveal that there exist a statistical connection between immigration and unemployment in Austria using yearly data for the period 1975-2012. The findings of the Johansen procedure show that immigration has a significant and positive effect on unemployment in the long-run. The vector error correction model indicates that immigration influence positively unemployment in the short-run. Furthermore, the finding of impulse response functions shows that the impact of unemployment on immigration, and vice versa, is insignificant, whereas the impact of real wages on per capita GDP only seems to be positive and significant, but, this effect dies off shortly.

Key words: immigration; labor market dynamics; co-integration.

Kurzfassung

Die vorliegende Arbeit untersucht die Beziehung zwischen Migrationsströmen und Arbeitsmarktvariablen, wie Arbeitslosigkeit, Reallohn und Pro-Kopf-Einkommen unter Verwendung vom Johansen-Kointegration-Verfahren, eines Fehlerkorrekturmodells sowie von Impulse-Antwort-Funktionen (IRFs). Die Ergebnisse der ökonometrischen Schätzung zeigen, dass es eine statische Relation zwischen Migration und Arbeitslosigkeit für Österreich zwischen 1975-2012 gibt. Das Kointegrationsverfahren zeigt, dass Migration langfristig einen signifikanten und positiven Effekt auf die Arbeitslosigkeit hat. Das Fehlerkorrekturmodell indiziert, dass Migration kurzfristig spürbare und positive Effekte auf die Arbeitslosigkeit hat. Nach den Ergebnissen der Impulse-Antwort-Funktion, führt Arbeitslosigkeit zu steigender Migration. Dieser Effekt ist jedoch insignifikant. Die Auswirkungen der Reallöhne auf Pro-Kopf-Einkommen sind lediglich positiv und signifikant. Der Effekt ist jedoch kurzfristiger Natur und scheint langfristig zu verschwinden.

Schlagwörter: Migration; Arbeitsmarktdynamik; Kointegration.

6

“Human migration is a complex and comprehensive issue which has to be understood as a consequence of political, ideological, social and economic process (Mückler, 2004; p.42)”.

1. INTRODUCTION

igration has become in recent years a hotly debated issue especially in

countries with high migration inflows. From the middle of the last century

taking the international migration flows into account one can realize that

immigration is a ‘matter’ of some industrialized Western European countries

(e.g., Austria, France, Germany, and UK), some Asian countries (e.g., Australia,

Japan) and North America (e.g., Canada and USA), that have attracted

increasing numbers of migrants from all of the world, primarily from Africa and

Latin America, marked as countries with negative net migration. In 2013, 232

million people - or roughly 3.4% of the world’s population - are foreign-born, i.e.

they reside in a country where they were not born. A majority of these

international migrants, 71%, originated in the South1. These considerable

migrant flows have consequently caused social, cultural and especially

economic changes both for the sending and receiving countries2 and have also

caused to conduct different migration research in different scientific disciplines

in order to deal with the general reasons, causes and impacts of immigration.

Austria is a traditional immigration country, marked by gradual migratory waves

in consequence of economic factors such as a gap in the labor supplies due to

economic booms in the 1960s and 1990s, the recession in the wake of the oil

1 See: UN International Migration Report, 2013. Note that the term “North” refers to countries or regions traditionally classified for statistical purposes as “developed,” while the term “South” refers to those classified as “developing.” The developed regions include Europe and Northern America plus Australia, New Zealand and Japan. 2 International migrants sent $529 billion in remittances back to their home Countries in 2012, according to the World Bank. The United States is the number one sender of international remittances, accounting for nearly a quarter of them (23.3%). India ranks at the top of all countries that receive remittances, with $69 billion on 2012. Moreover, about $4.9 billion in remittances were sent from Austria to other countries in 2012. Serbia ($1.9 billion), Germany ($782 million) and Czech Republic ($275 million) are the top three countries that receive remittances from Austria, respectively. Source: http://www.pewsocialtrends.org/2014/02/20/remittance-map/

M

7

crisis in 1974, and political factors such as the fall of the Iron Curtain in 1989,

the war in Yugoslavia of the earlier 1990s (BIFFL, 2004). As the fraction of the

migrants continuously increases, the role of immigrants and their contribution to

the economy become a key issue in Austria as well. One of the main arguments

for legally restricting immigration is the fear that immigrants will displace existing

native workers or that immigrants are less skilled, have lower chances in finding

jobs, and put pressure on the public purse. However, contrary to that, a range of

empirical studies on the impact of immigration on the local labor markets for

different countries either do not offer clear results or point to small effects of

migration so that the economic outcome is difficult to determine a priori.

Consequently, whether immigrants displace native workers or create jobs and

therefore are responsible for unemployment of the natives depends crucially on

the economic structure of a country such as size, openness, degree of labor

market flexibility3, degree of substitution between natives and immigrants, the

size of the immigrant inflows and the strictness of immigration policy, and on

relative abundance of natives in different skill, education, occupation and/or

experience groups, and immigrant’s role as producer in supply side and

consumer in demand side (GROSS, 1999).

Briefly, it can be argued that human migration and its effects on the overall

economy depend crucially on the viewpoint of the social sciences, on the

assessment and interpretation of different techniques, methodologies and

scientific theories. For this and some other reasons, this research paper adopts

a data-driven rather than a theory-driven approach. We estimate an

econometric model and we test for causal relationships among immigration,

economic growth, wages and especially unemployment, with long-term and

short-term distinction. We consider immigration effects on the Austrian labor

market solely from the viewpoint of the host country’s economy. Thus, the

overall research question of this article is the following: Do changes in Austrian

immigration rate cause changes in Austria’s unemployment? Is there a positive

relationship? If any, what are the short-run and long-run effects of immigration

3 “(..) Research on [wage] for European countries is rare and especially focused on Germany and Austria. As it is common knowledge that wages in Europe are more rigid than in the USA, European studies more often look at (un)employment effects of immigration and less at wage effects, (OKKERSE, 2008)”.

8

on the Austrian unemployment? Moreover, do increases in Austrian immigration

rate cause a decrease in Austrian workers’ wage?

The paper is organized as follows. The next section presents simple theoretical

background of the impact of immigration and surveys the literature on the

immigration impacts. In section III we describe Austrian immigration in detail

especially from 1960 on. In section IV we present the data used and the

descriptive analysis in order to illustrate our econometric time series model

using several statistic models by testing the relationship between

unemployment and immigration for the Austrian case. We finish with a

conclusion in section V.

2. THEORETICAL BACKGROUND AND LITERATURE REVIEW

After WW2, on the ground of the ‘economic miracle’ that caused a gap in

the labor supply, labor migration was permitted legally in most Western

European countries. As the flow of immigration increased, questions on the

labor market effects of immigration became a key topic for the literature of

economic science4. Moreover, the question of immigration and -in a wider

sense- its effect on countries’ economies has also sparked a debate especially

in politics thereby in society in these countries. One of the main arguments for

legally restricting immigration is the fear that immigrants might have a negative

impact on labor market prospects of natives. In other words, this fear arises

from the assumption that an increased immigration rate will increase the

unemployment rate and reduce the wages of natives and will thus activate the

displacement of domestic workers by immigrant workers. However, economic

theory and empirical evidence do no offer clear results for the overall

immigration effects. Consequently, whether immigrants cause unemployment of

native workers, depends crucially on the degree of labor market flexibility such

as wages; on the substitution or complementarity of native and immigrant

4 Why do some people migrate? And what happens when they do? Although these questions are important since they form the basis of different migration theories- such as non-economic factors, role of the social network for explaining of pull and push factors- we forego to go into detail, since this would exceed the scope of our work.

9

workers, on the size of the immigrant inflows and on the extent to which

immigrants' demand for goods and services generate more jobs than they

themselves occupy", (GROSS, 1999).

Theoretical aspects of labor market effects of immigration are usually

described using a neo-classical competitive model of supply and demand in the

market for labor services (see for instance JOHNSON, 1980; CHISWICK, 1982;

GREENWOOD & MCDOWELL, 1994). According to standard economic theory

an additional supply of workers that is an increase in immigration5 into an

economy is expected to reduce wages for native workers, since more

individuals are willing to supply their labor at a given wage. However, the effect

of immigrants on wages, especially, depends on whether immigrants and

natives are substitutes or complements. When natives and immigrants are

homogeneous in terms of their skill level (productivity) with the underlying

assumption that all labor enters the firm’s production function as a single input

and perfect substitutes in production an increase in the supply of immigration6 in

the short-run will lead to an increased competition in the same labor market for

both groups. As shown in Figure 1, this will reduce the wage for immigrant

workers (from W0 to W1) and increase total employment 7 (from N0 to E1). At

the lower wage, however, the number of natives who work declines (from N0 to

N1), since firms are expected to respond to immigration by hiring more

immigrant workers as they are cheaper than natives. Note that the magnitude of

these decreases depends on the size of the immigrant inflows and how

responsive labor supply and demand are to changes in wages: For any given

responsiveness of labor supply to changes in wages, the bigger the immigrant

inflow, the bigger the decreases in natives’ wages and employment

(ORRENIUS et al., 2013). As a result, for the short-run impact of immigration

5 Demand for labor is assumed to be a downward-sloping function of the wage, and also assumed to be unaffected by immigration. Moreover, it assumes an economy that is closed to trade with other regions so that an inflow of immigrants cannot lead to an increase in the production of traded goods or an in- or out-migration of resident labor. The increase in labor relative to capital is not allowed to simulate an inflow of capital or the adoption of new production techniques (CARTER et al., 2006). 6 Here it is assumed that the supply of immigrant labor is perfectly inelastic; that is, wages do not influence the amount of immigrant labor supplied. (CARTER et al., 2006). 7 Note that if there are restrictions of wage adjustment because of minimum wage or intervention of union for example then an increase in immigration will cause unemployment.

10

when immigrants and natives are perfect substitutes, an adverse effect on

native workers will occur (FERIDUN, 2007).

Figure 1: Short-Run Impact of Immigration when Immigrants and Natives are Perfect Substitutes

Source 1: Own representation based on BORJAS (2007).

The assumption that native workers and immigrants are perfect substitutes is

questionable. It might be that immigrant and native workers are not competing

for the same types of jobs where they are employed in two distinct labor

markets. Consequently, if immigrants and native workers are complements in

production, then presence of immigrants increases native productivity since

natives can now specialize in tasks that are better suited to their skills. If the two

groups are complements, an increase in the number of immigrants raises the

marginal product of natives, shifting up the demand curve for the native-born

workers. As shown in Figure 2, this increase in native productivity raises the

native wage (from W0 to W1). Moreover, some natives who previously did not

find it profitable to work now see the higher wage rates as an additional

incentive to enter the labor market, and native employment also raises from N0

to N1. As a result, for the short-run impact of immigration when immigrants and

natives are complements, a positive effect on native workers will occur

(BORJAS, 2007).

11

Figure 2: Short-Run Impact of Immigration when Immigrants and Natives are Complements

Source 2: Note that the labor market denotes here the supply and demand for native workers. Source: Own representation based on BORJAS (2007).

In the above discussion, product demand was assumed as fixed. However,

immigration has both demand and supply side effects in goods market that is

immigrants are not only competitors on the labor market for native workers

since it is possible that they increase productivity and income of the production

factors. Moreover, immigrants are also consumers demanding goods and

services produced by native firms and laborers. They can increase consumer

utility by putting a downward pressure on prices. When both demand and

supply effects are present, the net effect on the natives will depend on the

immigrants’ marginal propensity to consume and the chance of finding a job

relative to natives. If, for example, immigrants’ relative expenditure is less than

their relative employment, then the demand for labor will shift to a less extent

than the supply of labor and therefore some natives will lose their jobs

(FERIDUN, 2007).

There are large empirical studies that have examined the economic

(GDP, Economic Growth, Wage and [UN] Employment), demographic and fiscal

12

effects of immigration for various countries8. Given the above theoretical

predictions, most of the empirical evidence suggests that at the national level

the effects of immigration on (un)employment outcomes are minimal and that

immigrants are complements to natives but there is some evidence of wage

effects (BLANCHFLOWER et al., 2007). For example BORJAS (1994) finds that

immigrants have a small negative effect on the wage of native-born workers.

FRIEDBERG and HUNT (2005) have also shown that a 10% increase in

immigrants in the US and other advanced Western countries causes a fall in

wages of at most 1%. In another paper, BORJAS and KATZ (2005) have

shown that an 11% increase in the US labor force between the years 1980-

2000 resulted in an overall loss of about 3% of the real value of wages, and that

this loss reached 9% for high school dropouts. However, JAEGER (1995)

reported that increases in the immigrant share of the labor force during the

1980s accounted for 6% of the increase in the college/high-school wage

differential, and that immigration caused a 3 to 5% decrease in the wages of

high school dropouts in the aggregate of the 50 largest metropolitan areas.

Moreover, ALTONJI and CARD (1991) find significant wage effects during the

1970s and find furthermore an equally small positive effect suggesting the two

types of workers are complements rather than substitutes. CARD (2007) finds

also that a 10% increase in the immigrant’s share of workforce increased

average wages by 6%. On the contrary, LEMOS and PORTES (2008) have

failed to find any convincing evidence of an effect of A8 immigration either on

the average or at any point of the UK wage distribution9.

The impact of immigration can differ by countries, especially between US and

Europe there are important differences regarding the degree of the labor market

flexibility (wage flexibility and labor mobility). Countries with a higher degree of

labor market flexibility are more able to absorb immigrants without increases in

unemployment (RUHS, 2006). Lower labor market flexibility in Europe than in

the US could mean that immigration affects employment more than wages.

8 We will only highlight some of the conclusions of the earlier literature that are relevant for the current analysis. 9 However, one important limitation of the factor proportion approach is that it relies on outside of estimates of relative wage elasticizes to stimulate the impact of immigration on wages, DiNardo (1997). Thus, its validity depends on both the theoretical model and the estimate relative wage elasticizes to be correct.

13

Nevertheless, European studies tend to look at potential (un)employment

effects of immigration rather than at wage effects. Surprisingly, wage effects

found in European studies are more negative than wage effects found in the US

studies. Wage effects in the USA are at most −1.5% for a 1 percentage point

increase in immigrant share (GOLDIN, 1994). DE NEW & ZIMMERMANN

(1994a, b) find wage effects of − 3.3% and even −6.4% for a 1 percentage point

increase in immigrant share in Germany.

Empirical studies on the so-called displacement effect through

immigration were also stressed by different economists. For example, CARD

(2001) finds that, for an in increase in the U.S. immigrant share by 1%, the

native employment to population ratio would decrease by at most 1 percentage.

ANGRIST and KUGLER (2003) find larger effects for EU countries with a

maximal decrease of 1.6 percentage points. For the Austrian case, WINTER-

EBMER and ZWEIMÜLLER (2000) refer to an important point namely the

employment effect. In their study, they find no significant effect of immigration

on the probability of entering unemployment in Austria. However, this does not

mean that immigration had no employment effect at all; it only means that

employed workers were not affected. They show immigration has an impact on

the unemployed who find it more difficult to get back to work. When the

immigrant share increases by 1 percentage point, unemployment duration

increases by 5%. Moreover, BRANDEL (1994) pointed out an interesting

displacement effect of the surge of new immigrants during the 1990s in Austria

and find a significant displacement of guest-workers of earlier generations, but

also of natives: 60% of all firms in their sample with shrinking employment of

natives increased the employment of foreigners in the period 1989 to 199110.

Nevertheless, it should be noted that wage and employment effects and the

degree of substitutability of immigrants with native workers can change over

time.

Finally, studies of the aggregate relationship between immigration and

unemployment overwhelmingly reject the hypothesis of an adverse effect

(GROSS, 1999). For example, POPE and WITHERS (1993) investigate the

relationship between immigration and unemployment by applying both a 10 BORJAS (1994) argues that newly arrived immigrants are inherently different from those who migrated 20 years ago.

14

structural disequilibrium model and causality tests and found that immigrants to

Australia do not increase unemployment, especially, that the long-term effects

of immigration on unemployment are negligible. MARR and SIKLOS (1994),

examine the influence of past, current, as well as future immigration jointly on

the current unemployment rate in Canada, using quarterly data for the period

1992-1990. Their findings show that current increases in the unemployment rate

reduced future immigration rates before 1978. After 1978, however, there is a

positive association between past immigration and current unemployment.

Moreover, they show that the results are in line with those of POPE and

WITHERS (1993). In another study, MARR and SIKLOS (1995) examined the

relationship between immigration and unemployment in Canada using annual

data from 1926 to 1992. They conduct Granger-causality tests on

unemployment and immigration and an unrestricted VAR approach on

unemployment, immigration, wage and real GDP. The Granger-causality tests

showed that immigration was not caused by past unemployment, however, that

past immigration did cause unemployment. In other words, the causality runs

unidirectional from past immigration to unemployment. On the contrary, SHAN

et al. (1999) examine the relationship between unemployment and immigration

for Australia and New Zealand by using Vector Autoregression Model (VAR),

however, they find no such causality from immigration to unemployment.

GROSS (1998) finds a negative relationship between unemployment and

immigration in the long-run and a positive, but small, correlation in the short-run

in a regional market such as British Columbia. More recently, FERIDUN (2005)

examined the relationship between immigration, unemployment and economic

development in Norway using Granger causality tests. He finds a positive

relationship between immigration and GDP per capita and that immigration has

no impact on unemployment, and vice versa. Based on French data from 1970

to 2008 FROMENTIN (2013) examined the relationship between immigration,

labor market and economic development. Using Johansen’s co-integration

procedure, he determines long-run and short-run relationships in a VECM. He

finds a negative relationship between immigration and unemployment in both

the short and long term. That is, in the short run immigrants seem to displace

native workers or take away available jobs that are neglected by native workers.

15

In the long run, however, they seem to create more jobs through their demand

for goods and services. Moreover, the study also reveals that immigration has a

positive impact on the real wages, indicating that immigrants are more

complements rather than substitutes with natives.

3. IMMIGRATION TO AUSTRIA SINCE THE 1960s

Looking at Austria’s migration development reveals that, since WW2,

Austria has been simultaneously a country of temporal as well as permanent

immigration for both political, demographic and economic reasons. On the basis

of political crisis in the former communist countries as a result of the political

uprisings and repressions, Austria became a major destination for refugees

from these countries. Most of these refugees, however, stayed in Austria for a

short time only. In 1956, for example, over 180,000 Hungarian refugees came

to Austria but only about 20,000 of them were granted asylum and stayed in

Austria. The same pattern occurred for refugees from other countries: for

example due to the “Prague Spring” of 1968 and then to the crushing of the

“Solidarnosc” movement and the imposition of martial law in Poland in 1981

and 1982 about 162,000 Czechoslovakians and about 150,000 Poles came to

Austria respectively. Again the majority of these refugees returned either to their

home countries or travelled on to other Western countries (BIFFL, 2004; p.11).

3.1 Phase: From Emigration Country to Immigration Country (1960 – 1974) In 1960, and after, Austria started to recruit foreign workers, which on the

one hand trace back to the business cycle upswings of the 1950s in other

Western countries, where most of Austrian workers emigrated especially to

Germany and Switzerland and on the other hand due to the gap in the labor

supply, immigrants were supposed to reduce labor scarcities in Austria. The so-

called “Raab-Olah-Agreement”11 or “guest worker program”-an agreement

between firms and trade unions- was implemented in 1961. The first contract

was signed in 1962 with Spain, followed by Turkey in 1964 and two years later

by Yugoslavia. As Figure 3 shows, the first peak of the migration is achieved

11 The former WKO President Julius Raab and ÖBG President Franz Olah signed Raab-Olah Agreement in 1961.

16

with about 227,000 employed guest workers in 1973. Yugoslav nationals

formed the largest share of the guest workers with 78.5% in 1973, followed by

Turks with 11.8%. The main principle of the guest worker system was based on

“rotation”. The basic idea of policy-makers was to invite immigrants as cyclical

shock absorber that can be sent back again when the economic situation is bad

(BIFFL, 2004; p.10-11). However, this system never worked, because on the

one hand not enough workers were recruited: Austrian companies encouraged

their foreign employees to bring close relatives to Austria. On the other hand, in

order to cut costs, many companies were not willing to enroll new workers each

year12.

3.2 Phase: Between the Conflicting Priorities of Return and Settlement (1974–1988)

Due to the recession in the wake of the oil crisis in 1974 and the fact that

Austrians who had been working abroad returned home, a recruitment ban has

been enforced and the number of foreign workers begin to decrease (BIFFL,

2004; p.11). The Alien Employment Act of 1976 governs the preference for

Austrian workers over foreign workers. As a result, the share of foreign workers

in the Austrian labor force decreased about 40% between 1974 and 1984, but

the percentage of foreign residents remained almost at the same level, because

the number of returning immigrants was compensated by the incipient family

reunions (BIFFL, 2004; p.12).

3.3 Phase: Consequences of the end of Europe's division on immigration to Austria (1989-1993)

Between the period 1989 and 1993, with the dismantling of the Iron

Curtain, the war in Yugoslavia and the growing need for immigrants due to the

economic boom, another peak period of immigration began to start. According

to the Census 1991, even approximately 690,000 foreign nationals were living

12See:http://medienservicestelle.at/migration_bewegt/2011/05/25/neue-osterreichische-migrationsgeschichte/

17

in Austria, that is the number of foreigners living in Austria has almost doubled,

where the Austrian population was roughly 7.8 million13.

Figure 3: Foreign Population and Foreign Labor Force in Austria (1962-2012)

Source 3: Own representation based on Statistics Austria.

Moreover, the foreign share in the total population in Austria rose steadily from

approximately 1.5% in 1961 to more than 8 % in the early 1990s and topped the

10% mark for the first time in 2002 (see Figure 4). During the same period,

unemployment increased from 149,200 persons (thereof 10,000 foreigners) to

195,100 (27,100 foreigners). In 1989, the net immigration amounted to +64,600

immigrants and in the following three years it passed the level of 80,000 a year

(BIFFL, 2004; p.12).

13See:http://medienservicestelle.at/migration_bewegt/2011/05/25/neue-osterreichische-migrationsgeschichte/

0 100.000 200.000 300.000 400.000 500.000 600.000 700.000 800.000 900.000

1.000.000

1962

19

64

1966

19

68

1970

19

72

1974

19

76

1978

19

80

1982

19

84

1986

19

88

1990

19

92

1994

19

96

1998

20

00

2002

20

04

2006

20

08

2010

20

12

Foreign Population and Foreign Labor Force since 1962

Foreign PoP Foreign labor force

18

Figure 4: Foreign Share in the Total Population in Austria (1961-2012)

Source 4: Own representation based on Statistics Austria.

3.4 Phase: From the unregulated to regulated Migration (1994-until today) In 1990, the Austrian government tightened its immigration policy by

implementing a quota for work permits, which bound the maximum share of

foreign workers to the total labor force. In detail, a maximum of 10% of the total

work force could be foreign workers. Generally, the quotas vary from 8% to 10%

of the total work force. The Austrian government tightened its immigration policy

by implementing further series of new laws in 1992/199314. The guest workers

scheme was replaced by an annual quota system for new residence permits,

which reduced continuously the net immigration into Austria. Between 1993 and

2001, the net immigration amounted to 159,000 persons, which means that the

yearly net immigration did not exceed 20,000 persons during the 1990s (BIFFL,

2004; p.12). Figure 5 shows the net migration balance rate15 between 1975 and

2012 for Austria.

14 In 1993: A new Aliens Act and a new residence law was enforced. From this point, a separation between tourist visas and immigrant visas will be drawn. In the same year, the rightwing Freedom Party (FP) organized the "anti-foreigner referendum" with 416 531 signatures. Shortly thereafter, the "sea of lights" in which around 250,000 people carried protest against xenophobia. See: http://medienservicestelle.at/migration_bewegt/2011/05/25/neue-osterreichische- migrationsgeschichte/ 15 Difference of immigrants and emigrants of Austria divided per 1,000 inhabitants on an annual average.

0,0

2,0

4,0

6,0

8,0

10,0

12,0

14,0

1961

19

63

1965

19

67

1969

19

71

1973

19

75

1977

19

79

1981

19

83

1985

19

87

1989

19

91

1993

19

95

1997

19

99

2001

20

03

2005

20

07

2009

20

11

Foreign Share in the Total Population since 1961

19

Figure 5: Net Migration Balance Rate in Austria (1975-2012)

Source 5: Own representation based on Statistics Austria: “Wanderungen”

ASYLUM APPLICATIONS The period 1997-2002 for Austria was marked by huge increases in

asylum applications. The number of asylum applications has increased six-fold

and rose to almost 40,000 in 2002.

ACCESSION TO THE EU: With its accession to the European Union in 1995 and the two EU

enlargement rounds in 2004 and 2007, Austria has attracted increasing

numbers of migrants from the EU15 and the enlargement countries, many of

them were on a more temporary basis (MAYR et al., 2012)16 and caused to

break new records.

Today, Austria ranks among the EU15 countries with the highest share of

foreigners in the population: According to the Census of 2001, 12.5% of the

Austrian population was born in foreign countries. In 2012, on average some

1.579 million men and women (18.9% of the population) with foreign

background were living in Austria. 1.167 million of them were born abroad, while

412,000 people were descendants of foreign-born parents but born in Austria

and thus counted as “second generation”. In an international comparison, the 16 Seasonal workers, who are allowed to stay up to one year and re-apply after a two months break.

-4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9

10

Net Migration Balance Rate (1975-2012)

20

share of foreign-born persons is almost the same as that of the classic

immigration country USA, where, according to the 2010 Census, 12.9% of

inhabitants were born abroad17. Of the 1.579 million people with a migration

background over a third (35% or 547,000) came from another EU country,

another third (32 % or 512,000) were in the successor states of Yugoslavia

(excluding Slovenia, which has been 2004 EU member). Persons with a Turkish

background accounted for 17% (275,000 persons), while about 16% came from

people from other European countries and other parts of the world (246,000

people). In other words two-thirds of the immigrants come from non-EU

countries.

NATURALIZATIONS IN AUSTRIA Around two fifths of the Austrian resident population born abroad have

Austrian citizenship: for the most part they moved to Austria and were

subsequently naturalized. The largest group of Austrians born abroad (73,000

people) were those with a place of birth in Germany. Turkey comes second

(71,000 people), followed by Bosnia and Herzegovina (55,000), Serbia

(42,000), the Czech Republic (33,000), Romania (26,000) and Poland (21,000).

(STATISTICS AUSTRIA). As shown in Figure 6, in 2003 about 45,000 foreign

nationals were naturalized, which is also the absolute naturalization record for

Austria. However, since 2004, the number of naturalizations has declined and

fell to 2010 by more than four-fifths compared to the high record of 2003. The

sharp increase in the naturalization of national foreigners in 2003, however, has

not occurred due to a new more liberal policy, rather the new record has to do

primarily with the long lasting residence of migrants in Austria, who are normally

granted citizenship after 10 years (BIFFL, 2004; p. 14).

17 The reason for the high share of foreign nationals living in Austria, however, has crucially to do with the fact that Austria follows the so-called “ius-sanguinis” principle, where a foreign child receives the nationality of its parents, contrary to the “ius-soli” principle, where the place of birth is the decisive factor for obtaining the country’s nationality (BIFFL, 2004; p. 5).

21

Figure 6: Naturalized Persons in Austria (1960-2013)

Source 6: Own representation based on Statistics Austria: “Naturalisation”.

UNEMPLOYMENT IN AUSTRIA

Overall, around 35% of job-seeking people and around 20% of non-self-

employed foreign workers have a migration background in Austria18. The share

of foreigners in the Austrian labor force continues to rise, increasing from 9.1%

in 1991 to 13.4% in 2008 (that of the foreign-born was 16.3% in 2008) (MAYR

et al., 2012). Migrants, thus play an important role in the Austrian labor market.

Unemployment is on average typically higher among foreign workers than

among native workers in Austria. The unemployment rate varied between 7.8%

and 10.7% during 1990 - 201319 compared with a mean of 6.6% for natives in

the same period. As shown in Figure 7, the unemployment rates among natives

remain below 7.5%, while unemployment among foreign workers exceeded

10% mark for the first time in 2004.

18 Media Arbeiterkammer: “Arbeitsmarkt im Fokus-Arbeitsmarktanalyse des 1. Halbjahres 2013“. 19 Own calculation. Based on data of ELISweb.

0

5.000

10.000

15.000

20.000

25.000

30.000

35.000

40.000

45.000

50.000 19

60

1962

19

64

1966

19

68

1970

19

72

1974

19

76

1978

19

80

1982

19

84

1986

19

88

1990

19

92

1994

19

96

1998

20

00

2002

20

04

2006

20

08

2010

20

12

Naturalized Persons in Austria (1960-2013)

22

Figure 7: Unemployment Rates of Natives and Immigrants in Austria (1975-

2012)

Source 7: Own representation based on data base of ELISweb: “Entwicklung des AusländerInnnenarbeitsmarktes ab 1975“.

Moreover, as shown in Figure 8, between 1975 and 1981 the relationship

between the unemployment rate and immigration shows an inverse relationship

and from 1981 to 1989 a move in the same direction. The period between 1989

and 1993 with a structural break in 1991 shows that net immigration rate

exceeded unemployment rate for the first time. Moreover, the period between

1994 and 2002 is marked by an inverse relationship, where unemployment rate

reached its first maximum in 1998 with 7.2%. Consequently, these ups and

downs in both series point to both economic and political crises and thereby to

restrictive policy reactions to immigration.

0 1 2 3 4 5 6 7 8 9

10 11

Unemployment Rates of Natives and Immigrants (1975-2012)

NAT-UNEM MIG-UNEM

23

Figure 8: Net Migration Balance Rate and Unemployment Rate in Austria (1975-

2012)

Source 8: Own representation based on data base of ELISweb: “Entwicklung des AusländerInnnenarbeitsmarktes ab 1975“and Statistics Austria: “Wanderungen”.

INCOME BY CITIZENSHIP Immigrants in Austria have on average lower income than natives. As

reported in Figure 9, the total median monthly income by citizenship for wage

and salary earners in 2012, shows that Austrians (AUT) earned about 2,093 €,

whereas foreign nationals (FORNAT) earned about 1,649 €. Moreover, the

difference becomes clearer, when incomes of migrants in Austria by their

citizenship are compared: Germans (GER) earned on average 2,026 €, followed

by successor states and former Yugoslavs (EXYU) with 1,719 €, Hungarians

(HUN) with 1,633 €, former Czechoslovakia (CZESVK) with 1,530 €, Turks

(TUR) with 1,529 €, Polish (POL) with 1,481 € and Romanian (ROU) with 1,354,

respectively. Note that, between the period 2003 and 2012, the median

earnings of Germans and former Yugoslavs are above-average among foreign

nationals.

-4

-2

0

2

4

6

8

10

in %

Net Migration Balance Rate and

Unemployment Rate (1975-2012)

UNEM Migration

24

Figure 9: Income by Citizenship in Austria (2003-2012)

Source 9: Own representation based on ELISweb: “Einkommen nach Staatsbürgersaft- Zeitreihe“.

These differences in earnings can be explained such that migrants have on

average lower skills, working in low-pay sectors, are overrepresented in

temporary-jobs, and have lower career opportunities (MAYR et al., 2012). On

the other hand, one reason why Germans earn relatively much more than other

foreigner nationals in Austria, is that Germans face less language barriers,

which make the access to the labor market easier for them. Figure 10 shows

that immigrants in Austria are overrepresented in labor-intensive service sectors

such as manufacturing (18.9%), hotels and restaurants (16.3%), sale (15.7%)

and in constructions (12.1%), respectively, that are more prone to business

cycle fluctuations and so to the risk of to be unemployed (MAYR et al., 2012).

1200 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

in €

Income By Citizenship in Austria (2003-2012)

AUT Foreigners GER CZESVK EXYU

POL ROU HUN TUR

25

Figure 10: Foreign Share by Industry in Austria (2008-2013)

Source 10: Own representation and calculations based on Austrian Social Security Database (ASSD).

Another important reason for the above average incidence of unemployment of

migrants results from the fact that there exist an employment protection for

native workers, which trace back to the 1930s. Accordingly, a foreign worker is

the first to be laid off if the enterprise reduces its work force (BIFFL, 2004; p.

15).

Nevertheless, these differences in unemployment rates and earning between

immigrants and natives and the fact that immigrants have demographic effect in

enlarging the population (and thus the tax base) and in altering its age – (and

gender) composition, are also reflected in the welfare system. Consequently,

the effects of immigrants on fiscal policy, taxes, social security system,

unemployment benefit system, the public pension system and on social transfer

system can be beneficial or harmful for the welfare system (MAYR, 2004).

As noted above, “Austria is implementing a highly restrictive policy in the area of

labor migration, also when it comes to the different permits that one can be

granted and to the conditions to be fulfilled for the granting thereof“, (CIRCO

and VILICS 2003; p. 16). In principle an employer may only employ immigrants

if they hold an appropriate work permit. However, taking the above fact into

account, that two-thirds of the immigrants come from non-EU countries and that

different work permit laws hold for EEA citizens – these obtained free access to

26

the labor market with Austria's accession to the European Union in 1995, except

the countries that joined in the EU/EEA in 2004 and 2007- and non-EEA citizens

for which the work permit takes several years until to obtain free access to the

labor market20, all this shows that, since recently, Austria has given preference

to a EU countries first in its immigration policy, whereas citizens of the classical

source countries, (former) Yugoslavia and Turkey are relatively less preferred.

According to Statistics Austria, in 2013, about 158,000 Germans constitute a

share of almost 16% and represent the largest group of foreigners in Austria

followed by Turks (113,000) and Serbians (112,000)21.

4. DATA SOURCE AND DESCRIPTIVE STATISTICS

This study uses annual data for the period 1975 and 2012 for Austria

from BUREAU OF LABOR STATISTICS, BALI -an online database query

system of market information supervised by the APF team of FEDERAL

MINISTRY OF LABOR, SOCIAL AFFAIRS AND CONSUMER PROTECTION-,

UNITED NATIONS STATISTICS DIVISION OF THE DEPARTMENT OF

ECONOMIC AND SOCIAL AFFAIRS (DESA) and STATISTICS AUSTRIA.

Immigration, denoted by MIG, is represented by the net migration rate

(difference of immigrants and emigrants of Austria), divided per 1000

inhabitants published by Statistics Austria. Wages, denoted by WAGE, are

represented by the hourly compensation costs in U.S. dollars in manufacturing

published by Bureau of Labor Statistics. The unemployment rate, denoted by

UNEM, in Austria is reported by the Public Employment Service Austria (AMS).

It is based on the number of unemployed persons registered at AMS offices and

the number of employees recorded by the Main Association of Austrian

20 Restricted work permit: valid for one year and only for a specific worker, firm and workplace within the firm. General work permit: After one year legal work, it can be applied, which is valid for two years within a given federal state. Finally, immigrants who have been working for 3 years are entitled to an exemption certificate, which is valid for 5 years throughout Austria. Moreover, immigrants with a “settlement permit” that is issued after 5 years of residence have unrestricted access to the labor market. Workers from the New EU member’s states could get a “confirmation of free movement” that entailed a work permit after one year of legal employment of 5 years of uninterrupted settlement in a federal state. With 1 July 2011 replaced the Red-White-Red Card the hitherto existing quota system for immigration to Austria (MAYR et al, 2012). The map is based on a points system, rated the German ability, age, and special vocational training and offers qualified third-country workers and their family members a new single permit for working and settling permanently in Austria. 21 See: Statistik Austria, Statistik des Bevölkerungsstandes, Wanderungsstatistik.

27

Insurance Institution, published by BALI22. GDP per capita at current prices in

U.S dollars, denoted by GDP, published by United Nations Statistics Division of

the Department of Economic and Social Affairs.

Table 1 reports the summary statistics for observation (obs.), mean, standard

deviation (std. dev.), minimum (min), maximum (max), skewness and kurtosis of

the data collected. WAGE and GDP are log-transformed since they grow with

main aggregates, denoted by L. There are only 38 observations on UNEM,

MIG, LGDP and LWAGE. The skewness coefficients for our four series are

close to zero and the kurtosis coefficients are close to three, which indicates

that the normal distribution is close to normal.

Table 1: Descriptive Statistics

Source 11: Own calculation with the use of E-Views software

4.1 ECONOMETRIC MODEL

The econometric model builds on the general equilibrium framework described

in GROSS (1998, 2002, 2004), ISLAM (2007) and FROMENTIN (2013). In

order to answer the research questions we will the estimate the following

equation:

∑ 𝒚𝒕𝒊=𝟏 = 𝜶𝟎 + ∑ (𝜷𝒊𝒊=𝟏 𝒚𝒕−𝟏 + 𝜸𝒊𝒙𝒕−𝟏) + 𝜺𝒕, (1)

where 𝑦𝑡 is a vector consisting of net immigration rates, unemployment rates,

hourly wages, in order to measure aggregate domestic labor market conditions

and per capita GDP, in order to measure aggregate domestic economic activity,

𝛼0 is a vector of constant terms, 𝛽𝑖 and 𝛾𝑖 are all matrices of parameters of the

22 The number of unemployed and employed persons extracted from the Labor Force Survey (LFS) complies with the Labor Force Concept, which is based on the ILO definition of employment and unemployment.

28

endogenous and exogenous variables respectively, 𝒙𝒕 is a vector of exogenous

variables, and 𝜀𝑡 is an error term, where 𝜀𝑡~ 𝑖𝑖𝑑 𝑁 (0,Σ). Unemployment,

immigration, GDP and wage are endogenously determined in the long term, but

not necessarily in the short term. Variables of interest are symmetrically and

endogenously determined. Equation (1) accounts for the supply and demand

effects of immigration as well as feedbacks from wage and the labor supply to

determine the final impact of immigrations on the target place (ISLAM, 2007).

The aim is to identify the dynamics adjustment of the labor market in the

short run and in the long run. In order to examine the long-run dynamic

response of the labor market to immigration we firstly concentrate on the co-

integration procedure developed by Johansen (1998, 1991), where the

unemployment rate, hourly wage, real GDP per capita and immigration rates

are considered to be simultaneously determined. The method is first used to

determine the number of co-integrating vectors, then it also yields a (Vector)

Error Correction Model ((V)ECM) that maps the short-run dynamic response of

our four series (Gross, 1998).

4.2 UNIT ROOT TESTS

Before following our strategies, it is important to determine whether model

series are stationary23. It is common, that many economic and financial time

series exhibit trending behavior or non-stationarity in the mean so that the

problem of the so-called spurious regression occurs. Moreover, it is also

important to ensure that time series are integrated of same order. Especially, if

the time series are integrated of order 1, denoted by 𝐼(1), then co-integration

techniques can be used to model these long-run relations. Hence, pre-testing

for unit roots is often a first step in co-integration modelling (ZIVOT &

ANDREWS, 1992). Consequently, to test whether our time series are stationary

or not, we use Augmented Dickey-Fuller tests (1979, 1981) (ADF) and Philips

and Perron (1988) (PP) tests for the existence of a unit root. The ADF and PP

tests differ mainly in how they treat serial correlation in test regressions. As we

shall see later, the results of testing for stationarity are not changed using the

23 Stationarity is generally defined by the feature that moments such as mean and standard deviation do not change over time and do not follow any trends.

29

test statistics of ADF or PP. The null hypotheses for ADF test and PP test are

always that the variable has a unit root and is thus non-stationary, whereas the

alternative hypothesis is that the variable has roots outside the unit circle and is

thus stationary.

Both the ADF test and PP test can be used in three variants, which correspond

to three versions of the underlying regression models24:

a.) Model A: Without intercept and without trend.

b.) Model B: With intercept but without trend.

c.) Model C: With intercept and trend.

Table 2 shows the results of both ADF25 and PP t-statistic tests in levels as well

as the critical values of MacKinnon (1991) for rejection of the hypothesis of a

unit root. If the ADF and PP calculated statistics are larger than the McKinnon’s

critical value, then the null hypothesis is not rejected. In Table 2, we can see

that none of our four series are stationary in levels so that existence of unit root

is supported.

Table 2: Stationarity Test in Levels (ADF and PP)

Note: McKinnon’s critical value; the number of lags is determined with Schwarz Information Criterion. Source 12: Own calculation with the use of E-Views software.

A common trend removal or de-trending procedure is first-differencing (a period-

to-period change). First differencing is appropriate for 𝐼(1) time series: Put

another way, a time series 𝑟𝑡 is integrated of order one, if 𝑟𝑡 is non-

24 Please see Appendix A for further details. 25 Note that, an examination of our data indicates that the LGDP and LWAGE series contain a linear time trend, whereas MIG and UNEM do not contain a linear time trend.

30

stationary but the first difference 𝑟𝑡 − 𝑟𝑡−1 is stationary and invertible. Table 3

reports the result as follows: For the series UNEM, MIG, LGDP and LWAGE,

the null hypothesis of a unit root is rejected at the 1% levels. As a result, the

time series are 𝐼(1), so that they are difference stationary. However, if the time

series have unit root and the order of integration is identical, the time series

might be co-integrated. This means that we will test for co-integration or the

long-run relationship among the four time series.

Table 3: Stationarity Test in First-Differencing (ADF and PP)

Note: McKinnon’s critical value; the number of lags is determined with Schwarz Information Criterion. ∆ is the first-difference operator. Source 13: Own calculation with the use of E-Views software.

4.3 LONG-RUN RELATIONSHIPS AND COINTEGRATED VARIABLES

The finding that many macro time series may contain a unit root has spurred the

development of the theory of non-stationary time series analysis. Engle and

Granger (1987) pointed out that a linear combination of two or more non-

stationary series may be stationary. If such a stationary linear combination

exists, the non-stationary time series are said to be co-integrated. The

stationary linear combination is called the co-integrating equation and is

interpreted as long-rung equilibrium relationship among the variables. In order

to test for co-integration between immigration and three macroeconomic

variables, we use Johansen’s Approach to Co-integration (1988). The Johansen

procedure is useful for investigating co-integration of several 𝐼(1) time series,

contrary to the Engle-Granger two-stage procedure which runs into problems for

larger co-integrating ranks. Moreover, the Johansen procedure enables us to

estimate the co-integrating vectors and the entire ECM as well.

31

The multivariate Johansen co-integration procedure allows testing for the

number of co-integrating relationships and it provides standard error of long-run

relationships and adjustment coefficients.

The vector autoregressive model (VAR) is a natural starting point for the co-

integration analysis developed by Johansen:

For a general VAR of order 𝑝, denoted as VAR (p)26:

𝑦𝑡 = 𝐴1𝑦𝑡−1+. . . +𝐴𝑝𝑦𝑡−𝑝 + 𝐵𝑧𝑡 + 𝜀𝑡, (2)

The VAR can be formulated as the long-run strategy:

∆𝑦𝑡 = ∏𝑦𝑡−1 + ∑ Γ𝑖𝑝−1𝑖=1 ∆𝑦𝑡−𝑖 + 𝐵𝑧𝑡 + 𝜀𝑡, 𝑡 = 1,2 … ,𝑇 (3)

∏ =∑ 𝐴𝑖𝑝𝑖=1 − 𝐼 (4)

Γ𝑖 = −∑ 𝐴𝑗𝑝𝑗=𝑖+1 . (5)

where 𝑦𝑡 is a k-dimensional vector variable that is integrated of order one at

most [here: unemployment rate, immigration rate, log of hourly wages and log

of per capita GDP], ∆ is the first-difference operator, 𝐵 is coefficient matrix to be

estimated, 𝑧𝑡 is a vector of deterministic variables (constant, a linear trend, and

seasonal dummies), 𝜀𝑡 is a vector of innovations that is assumed to be white

noise Gaussian (𝜀𝑡~ 𝑖𝑖𝑑.𝑁 (0,Ω)). Note that if the coefficient matrix ∏ has full

rank, the vector process 𝑦𝑡 is stationary.

A result from linear algebra implies that, if the coefficient matrix ∏ has reduced

rank 0 < 𝑟 < 𝐾, then there exist 𝑘 𝑥 𝑟 matrices 𝛼 𝑎𝑛𝑑 𝛽 each with rank 𝑟 such

that ∏ = 𝛼𝛽′. The Granger representation theorem says that 𝛽′𝑦𝑡 is 𝐼(0). 𝑟 is

the number of co-integrating vectors (the co-integration rank) and each column

of 𝛽 is a co-integration vector. The elements of 𝛼 are known as the adjustment

parameters in the vector error correction model.

Johansen’s method is to estimate the ∏ matrix from an unrestricted VAR and to

test the restrictions implied by the reduced rank ∏. In order to determine the

number of co-integration vectors, the trace test (λ𝑡𝑟𝑎𝑐𝑒) and the maximum 26 Please see Appendix B for further details.

32

eigenvalue statistic (λ𝑚𝑎𝑥) can be used. The trace test tests the null hypothesis

of 𝑟 co-integrating vectors against the alternative hypothesis of 𝑘 co-integration

vectors, where 𝑘 is the number of endogenous variables, for 𝑟 = 0, 1, … 𝑘 − 1,

[H0(r) against HA(k)]. The alternative of 𝑘 cointegration relations corresponds to

the case where none of the series has a unit root. The maximum-eigenvalue

test tests the null hypothesis of 𝑟 co-integration vectors against the alternative

hypothesis that there are 𝑟 + 1 co-integrating relationships among the series,

[H0(r) against HA(r+1)], (HJALMARSSON and ÖSTERHOLM, 2010):

λ𝑡𝑟𝑎𝑐𝑒 (𝑟) = −𝑇∑ ln (1 − λ𝑖)𝑘𝑖=𝑟+1 (6)

λ𝑚𝑎𝑥 (𝑟, 𝑟 + 1) = −𝑇∑ ln (1 − λ𝑟+1)𝑘𝑖=𝑟+1 (7)

Here 𝑇 is the sample size and λ𝑖 is the i:th largest canonical correlation.

Table 4 reports the Johansen co-integration test results, for the trace statistic

and for the maximum eigenvalue statistic: The first and third columns show that

the first three eigenvalues are much larger than the last eigenvalues for both

tests. This suggests that there exist three co-integrated relations (r=3) at the 5%

level, that is, there are three linearly independent combinations of the non-

stationary variables that are stationary. As a result, the null hypotheses 𝑟 = 0,

𝑟 ≤ 1 and 𝑟 ≤ 2 can be clearly rejected.

Table 4: Johansen’s Trace Test and Maximum Eigenvalue Results

Note that E-Views displays the critical values forthe trace statistic reported by Osterwald-Lenum (1992), not those tabulated in Johansen and Juselius (1990). Source 14: Own calculation with the use of E-Views software.

33

By normalizing the coefficient of the series, estimation of the co-integration

relationship enables the long-term equations to be obtained using the Johansen

approach. Consequently, the three equation relations can be written as follows:

1) If there is one co-integrating relation (𝑟 = 1), then the co-integrating

equation has the following form27:

Vector 1I:

𝑈𝑁𝐸𝑀𝑡 = 𝑐𝑜𝑛𝑠𝑡. + 3.52𝑀𝐼𝐺𝑡 − 147.14𝐿𝑂𝐺𝑊𝐴𝐺𝐸𝑡 + 127.13𝐿𝑂𝐺𝐺𝐷𝑃𝑡 + 𝑍𝑡

(0.22) (33.45) (32.74) (8)

According to this result, an increase in immigration leads to an increase in

unemployment in the long-run.

2) If there are two co-integrating relations (𝑟 = 2), then the two co-integrating equations have the following form28:

Vector 1II:

𝑈𝑁𝐸𝑀𝑡 = 𝑐𝑜𝑛𝑠𝑡. +86.06𝐿𝑂𝐺𝑊𝐴𝐺𝐸𝑡 − 81.29𝐿𝑂𝐺𝐺𝐷𝑃𝑡 + 𝑍𝑡

(9.04) (8.86) (9)

Vector 2II:

𝑀𝐼𝐺𝑡 = 𝑐𝑜𝑛𝑠𝑡. + 66.43𝐿𝑂𝐺𝑊𝐴𝐺𝐸𝑡 − 59.37𝐿𝑂𝐺𝐺𝐷𝑃𝑡 + 𝑍𝑡

(9.84) (9.65) (10)

In equations 9 and 10, the variables LWAGE and LGDP are nearly proportional

to each other. According to the vector 1II high wages relative to per capita GDP

levels lead to a higher unemployment in the long-run, whereas according to the

vector 2II high wages leads to a higher immigration in the long-run, which is

coherent with the economic theory as well.

3) If there are three co-integrating relations (𝑟 = 3), then the three co-integrating equations have the following form29:

Vector 1III: 27 Standard error in parentheses. 28 Standard error in parentheses 29 Standard error in parentheses.

34

𝑈𝑁𝐸𝑀𝑡 = 𝑐𝑜𝑛𝑠𝑡. + 23.34𝐿𝑂𝐺𝐺𝐷𝑃𝑡 + 𝑍𝑡

(6.48) (11)

Vector 2III:

𝑀𝐼𝐺𝑡 = 𝑐𝑜𝑛𝑠𝑡. + 21.39𝐿𝑂𝐺𝐺𝐷𝑃𝑡 + 𝑍𝑡

(5.01) (12)

Vector 3III:

𝐿𝑂𝐺𝑊𝐴𝐺𝐸𝑡 = 𝑐𝑜𝑛𝑠𝑡. + 1.21 𝐿𝑂𝐺𝐺𝐷𝑃𝑡 + 𝑍𝑡

(0.07) (13)

According to vector 1III, vector 2 III and vector 3 III in equations 11, 12 and 13, an

increase in GDP per capita levels lead to higher unemployment, to higher

migration and to higher wage levels in the long-run, respectively.

4.4 SHORT-RUN RELATIONSHIPS

Engle and Granger (1987) have shown that co-integration implies the existence

of a vector error correction model (VECM). The VECM links the long-rung

equilibrium relationship implied by co-integration with the short-run dynamic

adjustment mechanism that describes how the variables react when they move

out of long-run equilibrium. In other words, we identify the short-run between

immigration flows and the regional labor market using VECM. Based on

Fromentin (2013) and Gross (2004), our four-equation-system is as follows:

∆𝑚 = 𝛼1 + 𝛽11𝑖𝑘𝑘

∆𝑢𝑡−𝑘 + 𝛽12𝑖𝑘𝑘

∆𝑤𝑡−𝑘 + 𝛽13𝑖𝑘𝑘

∆𝑚𝑡−𝑘 + 𝛽14𝑖𝑘𝑘

∆𝑦𝑡−𝑘

+ λ1𝑖𝑒𝑐𝑚𝑡−1 + 𝑒𝑡

∆𝑢 = 𝛼2 + 𝛽21𝑖𝑘𝑘




∆𝑦𝑡−𝑘

+ λ2𝑖𝑒𝑐𝑚𝑡−1 + 𝑢𝑡

∆𝑤 = 𝛼3 + 𝛽31𝑖𝑘𝑘




∆𝑦𝑡−𝑘

+ λ3𝑖𝑒𝑐𝑚𝑡−1 + 𝑣𝑡

35

∆𝑦 = 𝛼4 + 𝛽41𝑖𝑘𝑘

∆𝑈𝑡−𝑘 + 𝛽42𝑖𝑘𝑘



∆𝑦𝑡−𝑘

+ λ4𝑖𝑒𝑐𝑚𝑡−1 + 𝑠𝑡

where, 𝑡 = 1, 2, … ,𝑇, ∆ corresponds to the first differences, 𝑚,𝑢,𝑤 𝑎𝑛𝑑 𝑦 are

the endogenous variables, 𝛼1,𝛼2,𝛼3 𝑎𝑛𝑑 𝛼4 correspond to the intercepts for

each variables, 𝑒𝑐𝑚𝑡−1 is a one-period lagged error correction term,

𝑒𝑡 ,𝑢𝑡,𝑣𝑡 𝑎𝑛𝑑 𝑠𝑡 are serially uncorrelated error terms. Note that under the

hypothesis of co-integration all the variables in the ECM model are I(0).

Table 5: Results of Adjustment Coefficients from Normalized Co-Integrating Vectors:

Standard errors in parentheses. ∆ corresponds to the first-differences.

According to the results in Table 5 - if there is one co-integrating relation (𝑟 = 1)

is assumed - then in the short-run, especially, unemployment shows a

significant reaction: Migration increases, the vector UNEM-3.5MIG decreases

(Equation 8), that is changes in unemployment rate (∆UNEM) react positively

only, other variables have little power of reaction. Moreover, about 8.6% (-

0.086) of disequilibrium is corrected each year by changes in unemployment

rate.

36

Furthermore, for the short-run relationship the estimates of adjustment

coefficients for vector 1II in Table 5 show that, wages significantly corrects the

disequilibrium by 10.8% (0.108) each year by changes in unemployment rates.

For vector 2II only unemployment reacts to changes where about 30.2% (0.302)

of disequilibrium is corrected each year by changes in immigration rates.

Figure 11: Impulse Response of Unemployment Rate on Migration Rate

Source 15: Own Calculation with the use of E-Views software: "Response to Cholesky ONE S.D Innovations +/- 2 S.E". Note that the impulse response function is statistically significant when both standard-error bands are above or below zero on the y-axis. Note also that in E-Views we selected an unrestricted VAR model with r=4.

4.5 IMPULSE RESPONSE FUNCTIONS

Due to the difficulty of interpreting the estimated coefficients for the VAR model,

it is common to summarize the results by means of response to shocks. Figure

11 shows the response of the unemployment, immigration, real wages and per

capita GDP levels to shocks in each variables. In particular, a shock in each

variable is allowed to affect all four series, with the effect of the shock having

37

both short-run and long-run impacts. Using a Cholesky decomposition on the

VAR model30, the impulse response functions in Figure 11 show that a one

standard shock in the change in the unemployment rate positively and

significantly impacts future changes in the unemployment rate in the short-run.

In the long-run, however, the effect seems to be insignificant (‘Response of

UNEM to UNEM’)31.

Figure 11 also shows that a one standard deviation shock to immigration

increases (insignificantly) unemployment both in the short-run and in the long-

run: Unemployment reaches a maximum about 3 years after the initial net

migration rate shock to the economy (‘Response of UNEM to MI’).

Furthermore, as shown in Figure 11, real wages and per capita GDP are nearly

proportional to each other. According to that a shock in the change in

immigration positively and insignificantly impacts both future changes in real

wages and in per capita GDP levels (‘Response of LWAGE to MI’ and

‘Response of LGDP to MI’).

Also shown in Figure 11 the impact of a shock in the change in the

unemployment rate negatively and insignificantly impact future changes in

immigration both in the in the short-run and in the long-run. (‘Response of MI to

UNEM’). The finding is in line with previous empirical studies (Pope and

Withers, 1985; Marr and Siklos, 1994 and Islam, 2007) as well. Moreover, the

impulse response function in Figure 11 shows also that immigration responds

positively and insignificantly to real wages in the short-run, but, this effect dies

off shortly and has no long-run effect (‘Response of MI to LWAGE’).

Summarized it can be said that only the impact of real wages on per capita

GDP is positive and significant, but again, this effect dies off shortly (‘Response

of LGDP to LWAGE’).

5. CONCLUSIONS

30 Note that due to the yearly frequency of the data, a 10-year period after the occurrence of the shocks is used for the analysis. 31 This also true for (‘Response of MI to MI), (‘Response of LAWAE to LWAGE) and (‘Response of LGDP to LGDP’)

38

This paper investigated the long-run and the short-run impacts of immigration to

the Austrian labor market variables, such as unemployment rate, wage level

and per capita GDP by using Johansen co-integration tests, a vector error

correction model, and impulse response functions, respectively. The estimation

results reveal that there exist a statistical connection between immigration and

unemployment in Austria using yearly data for the period 1975-2012. The

Johansen co-integration tests, through the statistics of the trace and of the

maximum-eigenvalue, revealed the existence of three co-integration vectors32.

The findings of the Johansen co-integration procedure show that the

immigration rate has a significant and positive effect on the unemployment rate

in the long-run in Austria. That is an increase in immigration leads to an

increase in unemployment in the long-run. Contrary to this result, impulse

response function, shows that the impact of immigration on the unemployment

rate is positive but insignificant both in the short-run and in the long-run.

Furthermore, the findings of the Johansen co-integration procedure induce also

that, a higher level of wages relative to GDP per capita levels leads both to a

higher level of unemployment and to immigration in the long-run, which is in line

with the economic theory as well.

The vector error correction model (VECM) indicates that the changes in

unemployment rate react significantly and positively to the immigration shocks

in the short-run, whereas both wage levels and per capita GDP react little to the

immigration shocks. One possible explanation of the positive impact of

immigration on unemployment in the short-run - as emphasized by OKKERSE

(2008) - could be attributed to an increase in job search time rather than to

displacement of native workers. Moreover, contrary to wage levels and GDP per

capita, the variations of unemployment rates have significant and positive

reactions to shocks in the short-run.

Furthermore, the finding of impulse response functions shows that the impact of

unemployment on immigration, and vice versa, is insignificant, whereas the

impact of real wages on per capita GDP seems to be positive and significant,

but, this effect dies off shortly.

32 Note that the interpretation of the results depend crucially on the number of co-integration relations and its difficulty increases the more co-integrating ranks exist.

39

6. REFERENCES

Altonji, J., & Card, D. (1991). The effects of immigration on the labor market outcomes of less-skilled natives. In J. Abowd and R. Freeman (eds), Immigration, Trade, and the Labor Market (pp.201-234). Chicago: IL: University of Chicago Press.

Angrist, J., & Kugler, A. (2002). Protective or Counter Productive? Labr Market Institutions and the Effect of Immigration on EU natives. The Economic Journal. 113: 302-331.

Austrian Social Security Database (ASSD). (n.d.). Retrieved from http://iambweb.ams.or.at/ambweb/

Biffl, G. (2004). The impact of immigration on Austria's Society. Vienna: A survey of recent Austrian migration research.

Blanchflower, D., Saleheen, J., & C., S. (2007). The Impact of Recent Migration from Eastern Europe on the UK Economy. UK: Research paper based on speech at te Cambrigeshire Chamber of Commerce, 4th January.

Boeri, T., & Brücker, H. (2005). Migration, Co-ordination Failures and EU Enlargement. Germany: IZA. Discussion Paper No. 1600.

Borjas, G. (1994). The Economics of Immigration. Journal of Economic Literature, Vol.32, No.4, pp. 1667-1717.

Borjas, G. (1996). The New Economics of Immigration. Affluent Americans Gain; Poor Americans Lose. Retrieved from THE Atlantic Online: http://www.theatlantic.com/past/docs/issues/96nov/immigrat/borjas.htm

Borjas, G. (2007). Labor Economics, Sixth Edition. McGraw-Hill Internation Edition 2013.

Borjas, G., & Katz, L. (2005). The Evolution of the Mexican-born workforce in the United States. NBER Working Paper. No. 11281.

Borjas, G., Freeman, R., & Katz, L. (1992). On the labor market effects of immigration and trade. In g.Borjas and R.B Freeman (eds), Immigration and the Work Force: Economic Consequences for the United States and Source Areas (pp.213-214). Chicago: IL: National Bureau of Economic Research, University of Chicago Press.

Brandel, F. H. (1994). Verdrängungsprozesse am Arbeitsmarkt. Vienna: Research Memorandum No. 345, Institute for Advanced Studies.

Card, D. (2001). Immigrant Inflows, Native Outflows, and the Local Market Impacts of higher Immigration . Journal of Labor Economics. 19: 22-64.

Card, D. (2007). How Immigration affects U.S. Cities. Center for Research and Analysis of Migration, Discussion Paper Series CDP No. 11/07.

40

Carter, S., & R., S. (2006). Labor Market Flooding? Migrant Destination and Wage Change during America's Age of Mass Migration. Manuscript.

Chiswick, B. (1982). The impact of immigration on the level and distribution of economic well-being. Washington DC: American Enterprose Institue for Public Policy Research.

De New, J., & Zimmermann, K. (1994a). Blue collar labor vulnerability: wage impacts of migration. In G.Steinmann and R.E Ulric (eds), The Economic Consequences of Immigration to Germany (pp. 81-99). . Heidelberg: Physica.

De New, J., & Zimmermann, K. (1994b). Native wage impacts of foreign labor: a random effects panel analysis. Journal of Population Economics 7: 117-192.

DiNardo, J. (1997). Comments and Discussion. Broking Papers on Economic Activity 1: 68-76.

Dvorak, J. a. (2011). Staat-Migration-Globalisierung. Wien: Facultas.

ELISweb- Wirschafts-und Arbeitsmarktinformationssystem. "Entwicklung des AusländerInnenarbeitsmarktes seit 1975". (n.d.). Retrieved from Ein Service des APF-Teams der Sektion VI/6 im Bundesministerium für Arbeit, Soziales und Konsumentenschutz.: http://www.dnet.at/elis/Arbeitsmarkt.aspx

Engle, R., & Granger, C. (1987). Co-integration and error correction: representation, estimation and testing. Econometrica. 55:251-276.

Feridun, M. (2005). Investigating the Economic Impact of Immigration on the Host Country: the Case of Norway. Journal of Economic Perspectives. 9: 23-44.

Feridun, M. (2007). Immigration, Income and Unemployment: An Application of the Bounds Testing Approaoch to Cointegration. Journal of Developing Areas, 41(1), 37-51.

Friedberg, R. M., & Hunt, J. (2005). The Impact of Immigrants on Host Country Wages, Employment and Growth. Journal of Economic Perspectives, 9(2): 23-44.

Fromentin, V. (2013). The Relationship between Immigration and Unemployment: The Case of France. France: Economic Analysis and Policy. Vol. 43, p. 51-66.

Goldin, C. (1994). The political economy of immigration restriction in the United States, 1890 to 1921. In G.Goldin and G. Libecap (eds), The Regulated Economy: An Historical Analysis of Political Economy (pp.223-257). Chicago: IL: University of Chicago Press.

41

Greenwood, M., & McDowell, J. (1994). The national labor market consequences of US immigration. In H.Giersch (ed.), Economic Aspects of Internatinal Migration (pp. 155-194). Heidelberg: Springer.

Gross, D. (1998). Immigration Flows and Regional Labor Market Dynamics. . IMF Working Paper.

Gross, D. (1999). Three Million Foreigners, Three Million Unemployled? Immigration and the French Labor Market. Working Paper, International Monetary Fund.

Gross, D. (2002). Financial Intermediation: A Contributing Factor to Economic Growth and Employment. ILO Working Papers 351412, International Labour Organization.

Gross, D. (2004). Impact of Immigrant Workers On A Regional Labour Market. Applied Economic Letters. 11: 405-408.

Hjalmarsson, E., & Österholm, P. (2010). Testing for Cointegration Using the Johansen Methodolog when Variables are Near-Integrated. IMF Working Paper. WP/07/141.

Islam, A. (2007). Immigration, Unemployment Relationship the Evidence from Canada. Australian Economic Papers. 46: 52-66.

Jaeger, D. (1995). Skill differences and the effect of immigrants on the wages of natives. US Bureau of Labor Statistics Economic Working Paper No. 273.

Johansen, S. (1988). Statistical analysis of cointegrating vectors. Journal of Economic Dynamics and control. 12:231-254.

Johansen, S. (1991). Estimation and hypothesis testing of cointegration vectors in Gaussian vectors autoregressive models. Econometrica. 59: 1551-1580.

Johnson, G. (1996). The labor market effects of immigration. Industrial and Labor Relations Review 33(3): 331-341.

Lemos, S., & Portes, J. (2008). New Labour? The Impact of Migration from Central and Eastern European Countries on the UK Labour Market. IZA Discussion Papers 3756. Institute for the Study of Labor (IZA).

MacKinnon, J. (1991). "Critical Values for Cointegration Tests," in Long-Run Economic Relationships, Readings in Cointegration, eds. R.F Engele nad C.W.J Granger . New York: Oxford University Press, pp. 266-276.

Marr, W., & Siklos, P. (1994). The Link Between Immigration and Unemployment in Canada. Journal of Policy Modeling 16(1): 1-25.

Marr, W., & Siklos, P. (1995). Immigration and Unemployment: A Canadian Macroeconomic perspective, in : D.J. DeVoetz (ed.), Diminishing

42

Returns: The economics of Canada's recent Immigration Policy. Toronto: The C.D: Howe Institute and the Laurier Institution: 293-330.

Mayr, K. (2004). The Fiscal Impact of Immigrants in Austria- A Generational Accounting Analysis . Linz: Working Paper No. 0409.

Mayr, K., & Prean, N. (2012). Unemployment of immigrants and natives over the business cycle: evidence from the Austrian labor market. Austria: NORFACE Migration Discussion Paper 2012-19 and NRN: The Austrian Center for Labor Economics and the Analysis of the Welfare State, Working Paper 1210.

Media Arbeiterkammer: Arbeitsmarkt im Fokus. Arbeitsmarktanalyse des 1. Halbjahres 2013. (2013). Retrieved from http://media.arbeiterkammer.at/PDF/Arbeitsmarkt_im_Fokus_1_2013.pdf

Medien-Serviecestelle. Neue ÖsterreicherInnen. Das Portal für JournalistInnen zu Migration und Integration. (2001, 5 25). Retrieved from http://medienservicestelle.at/migration_bewegt/2011/05/25/neue-osterreichische-migrationsgeschichte/

Mückler, H. (2004). Migrationsdynamiken: Auslöser, Erklärungsmodelle, Konsequenzen. WUV-Univ.-Verl.

Okkerse, L. (2008). How to Measure Labour Market Effects of Immigration: A Review . Journal of Economic Surveys. 22: 1-30.

Orrenius, P., & Zavodny, M. (2013). Immigrants in the U.S. Labor Market. USA: Federal Reserve Bank of Dallas. Research Depertment. Working Paper 1306. .

Osterwald-Lenum, M. (1992). A Note with Quantiles of the Asymptotic Distribution of the Maximum Likelihood Cointegration Rank Test Statistics. Oxford Bulletin of Economics and Statistics 54(3): 461-472.

Pope, D., & Withers, G. (1985). Immigration and Unemployment. Economic Record 61: 554-563.

Pope, D., & Withers, G. (1993). Do migrants rob jobs? Lessons of Australian history, 1861-1991. Journal of Economic History 53(4): 719-742.

Ruhs, M. (2006). Greasing the Wheels of the flexible labour market: East European Labour Immigration in the UK. Centre on Migration, Policy and Society. Working Paper No.38, University of Oxford.

Shan, J., Morris, A., & Sund, F. (1999). Immigration and Unemployment: New Evidence from Australia and New Zealand. . International Review of Applied Economics 13(2): 253-260.

Statistik Austria: "Wanderungen (Zuzüge und Wegzüge)". (n.d.). Retrieved from www.statistik.at/web_de/statistiken/bevoelkerung/wanderungen/index.html

43

Statistik Austria: Einbürgerungen seit 1946 nach Bundesländern. (n.d.). Retrieved from http://www.statistik.at/web_de/statistiken/bevoelkerung/einbuergerungen/index.html

United Nations, U. (2013). Department of Economic and Social Affairs - Population Division (2013). International Migration Report 2013.

Winter-Ebmer, R., & Zweimüller, J. (1996). Immigration and the earnings of young native workers. Oxford Economic Papers 48: 473-491.

Winter-Ebmer, R., & Zweimüller, J. (1999). Do Immigrants Displace Young Native Workers: The Austrian Experience. Austria: Journal of Population Economics, Vol. 12, 1999, 327-340.

Winter-Ebmer, R., & Zweimüller, J. (2000). Consequences of trade creation and increased immigration for the Austrian labour market. In M.Landesmann and K. Pichelmann (eds), Unemployment in Europe (pp.250-267). Basingstoke: Macmillan.

Zivot, E., & Andrews, D. (1992). Further Evidence on the Great Crash, the Oil-Price Shock, and the Unit-Root Hypothesis. Journal of Business and Economic Statistics. Vol.10, No.3, p. 251-270.

7. APPENDIX A

Unit Root Tests:

For unit root test we use augmented Dickey-Fuller (ADF) test and Phillips-Perron (PP) test.

1. The Augmented Dickey-Fuller (ADF) Test Consider first an AR(1) process, and for the moment let’s ignore the deterministic trend: 𝑦𝑡 = 𝜇 + 𝑝𝑦𝑡−1 + 𝜀𝑡, (1.1) where 𝜇 is a constant and 𝑝 is a coefficient parameter. 𝜀𝑡 is assumed to be white noise: (𝜀𝑡~ 𝑖𝑖𝑑.𝑁 (0,Ω)) and 𝑦 is a stationary series if −1 < 𝑝 < 1. The unit root test is then carried out for the null hypothesis 𝑝 = 1 against the one-sided alternative hypothesis of 𝑝 < 1:

𝐻0:𝑝 = 1 𝑣𝑠 𝐻1:𝑝 < 1

44

Note that PP tests take also the unit root as the null hypothesis. Subtracting 𝑦𝑡−1 from both sides of the equation (1) yields: 𝑦𝑡 − 𝑦𝑡−1 = 𝜇 + 𝑝𝑦𝑡−1 − 𝑦𝑡−1 + 𝜀𝑡 (1.2)

∆𝑦𝑡 = 𝜇 + 𝑦𝑡−1 (𝑝 − 1) + 𝜀𝑡 (1.3)

where ∆ is the first-difference operator. Using (𝑝 − 1) = 𝛾, we get:

∆𝑦𝑡 = 𝜇 + 𝛾𝑦𝑡−1 + 𝜀𝑡 (1.4)

Thus, the null and alternative hypotheses are

𝐻0: 𝛾 = 0 𝑣𝑠 𝐻1: 𝛾 < 0

The null hypothesis is that the variable contains a unit root ("𝑦𝑡 is a random walk”) and implies that the variables are non-stationary. The alternative hypothesis is that the variable does not contain unit root (“𝑦𝑡 is a stationary AR(1)”). Under the null, 𝑦𝑡 is first order integrated: 𝐼(1), because it is stationary after one difference. Under the alternative it is a mean reverting AR(1).

Moreover, the t-statistic under the null hypothesis of a unit root does not have the conventional t-distribution. For this reason MacKinnon (1991) critical values for unit root tests are taken into account. Notice that the test is left-tailed.

When testing for unit roots, it is crucial to specify the null and alternative hypotheses appropriately to characterize the trend properties of the data at hand. Thus, whether the null hypothesis includes a drift term and whether the regression used to obtain the test statistic includes a constant term and time trend.

Two other cases of test equation:

a.) Without intercept and without trend:

𝑦𝑡 = 𝑝𝑦𝑡−1 + 𝜀𝑡

𝑦𝑡 − 𝑦𝑡−1 = 𝑝𝑦𝑡−1 − 𝑦𝑡−1 + 𝜀𝑡

∆𝑦𝑡 = (𝑝 − 1)𝑦𝑡−1 + 𝜀𝑡

∆𝑦𝑡 = 𝛾𝑦𝑡−1 + 𝜀𝑡

Thus, the null and alternative hypothesis of ADF t-test is

𝐻0 :𝑝 = 1 ↔ 𝐻0 ∶ γ = 0

𝐻1:𝑝 < 1 ↔ 𝐻1 ∶ γ < 0

The null hypothesis is that the variable contains a unit root and implies that the variables are non-stationary.

b.) With intercept and trend.

45

𝑦𝑡 = 𝜇 + 𝑝𝑦𝑡−1 + λ𝑡 + 𝜀𝑡

𝑦𝑡 − 𝑦𝑡−1 = 𝜇 + 𝑝𝑦𝑡−1 − 𝑦𝑡−1 + λ𝑡 + 𝜀𝑡

∆𝑦𝑡 = 𝜇 + (𝑝 − 1)𝑦𝑡−1 + λ𝑡 + 𝜀𝑡

∆𝑦𝑡 = 𝜇 + 𝛾𝑦𝑡−1 + λ𝑡 + 𝜀𝑡

Thus, the null and alternative hypothesis of ADF t-test is

𝐻0 :𝑝 = 1 ↔ 𝐻0 ∶ γ = 0

𝐻1:𝑝 < 0 ↔ 𝐻1 ∶ γ < 0

The null hypothesis is that the variable contains a unit root and implies that the variables are non-stationary. Under the alternative it is trend stationary.

2. Phillips-Perron (PP) Test:

Consider first an AR(1) process, and for the moment let’s ignore the deterministic trend:

𝑦𝑡 = 𝛼 + 𝛽𝑦𝑡−1 + 𝜀𝑡, (2.1) where 𝛼 is a constant parameter and 𝛽 is a coefficient parameter. 𝜀𝑡 is assumed to be white noise: (𝜀𝑡~ 𝑖𝑖𝑑.𝑁 (0,Ω)) and 𝑦 is a stationary series if −1 < 𝛽 < 1. 𝑦𝑡 = 𝛼 + 𝛽𝑦𝑡−1 + 𝜀𝑡 (2.2)

𝑦𝑡 − 𝑦𝑡−1 = 𝛼 + 𝛽𝑦𝑡−1 − 𝑦𝑡−1 + 𝜀𝑡 (2.3)

∆𝑦𝑡 = 𝛼 + 𝛽𝑦𝑡−1 + 𝜀𝑡 (2.4)

In (2.4), 𝜀𝑡 might be autocorrelated. The Phillips-Perron (PP) Test uses a non-parametric method of controlling for higher-order serial correlation. In this sense, the PP test statistic is a Dickey-Fuller statistic that has been made robust to serial correlation by using the Newey-West autocorrelation consistent covariance matrix estimator:

𝑤2 = 𝛾0 + 2∑ (1 − 𝑣𝑞+1

)𝛾𝑗𝑞𝑣=1 ,

where

𝛾 = (∑ 𝜀𝑇𝑡=𝑗+1 𝜀−𝑗)/T

46

and q is the truncation lag. The PP t-statistic is computed as

𝑡𝑝𝑝 = 𝛾01/2𝑡𝑏𝑤

−(𝑤2 − 𝛾0)𝑇𝑠𝑏

2𝑤𝑠

where 𝑡𝑏, 𝑠𝑏 are the t-statistic and standard error of 𝛽 and 𝑠 is the standard error of the test regression. The critical values are the same as for the Dickey-Fuller statistic and again we use McKinnon’s critical values.

8. APPENDIX B

When time series variables are non-stationary, it is interesting to see if there is a certain common trend between those non-stationary series. If two non-stationary series 𝑋𝑡~𝐼(1),𝑌𝑡 ~𝐼(1) have a linear relationship such that 𝑍𝑡 = 𝑚 +𝛼𝑋𝑡 + 𝛽𝑌𝑡, (𝑍𝑡 is stationary), then we call the two series 𝑋𝑡 and 𝑌𝑡 are co-integrated. Thus, Johansen’s Approach to co-integration (1987) is used.

The two time series are summarized to a vector of 𝑦𝑡 and transferred in a VAR model.

𝑦𝑡 = 𝐴𝑦𝑡−1 + 𝜃 + 𝜀𝑡, (1)

where 𝑦𝑡 is a 𝑘 𝑥 1 vector of endogenous variables, A is the 𝑘 𝑥 𝑘 matrix of coefficients, 𝜃 is a 𝑘 𝑥 1 vector of constants and 𝜀𝑡 is a 𝑘 𝑥 1 of error terms that are assumed to be white noise Gaussian (𝜀𝑡~ 𝑖𝑖𝑑.𝑁 (0,Ω)).

Matrix notation:

𝑦𝑡 = 𝑦𝑡1

𝑦𝑡2, 𝑦𝑡−1 = 𝑦𝑡−1

1

𝑦𝑡−12 ,𝐴 =

𝑎11 𝑎12𝑎21 𝑎22,𝜃 = 𝜃0

1

𝜃02,𝜀𝑡 = 𝜀𝑡

1

𝜀𝑡2. (2)

The VAR model contains for each row a separate equation and can be easily generalized to higher lags 𝑝. A p-th order VAR, denoted 𝑉𝐴𝑅(𝑝), is

𝑦𝑡 = 𝐴1𝑦𝑡−1+. . . +𝐴𝑝𝑦𝑡−𝑝 + 𝜃 + 𝜀𝑡, (3)

It is also possible to take real exogenous repressors as vector 𝑥𝑡 into the VAR model;

𝑦𝑡 = 𝐴1𝑦𝑡−1+. . . +𝐴𝑝𝑦𝑡−𝑝 + 𝐵𝑥𝑡 + 𝜀𝑡, (4)

where 𝑦𝑡 is a vector of non-stationary 𝐼(1) variables, 𝐴1 − 𝐴𝑝,𝐵 are 𝑘 𝑥 𝑘 matrices of coefficients to be estimated and 𝑥𝑡 is a vector of deterministic variables (that may contain a constant, a linear trend and seasonal dummies).

47

A VAR model can be rewritten as a Vector Error Correction Model (VECM) (trick 𝑦𝑡 = 𝑦𝑡−1 + ∆𝑦𝑡) . For a VAR of order two, denoted as VAR (2), it can be formulated:

𝑦𝑡 = 𝐴1𝑦𝑡−1 + 𝐴2𝑦𝑡−2 + 𝐵𝑥𝑡 + 𝜀𝑡 |+ 𝐴2𝑦𝑡−1- 𝐴2𝑦𝑡−1 (5)

𝑦𝑡 = 𝐴1𝑦𝑡−1 − 𝐴2(𝑦𝑡−1 + 𝑦𝑡−2) + 𝐴2𝑦𝑡−1 + 𝐵𝑥𝑡 + 𝜃 + 𝜀𝑡 (6)

𝑦𝑡 = (𝐴1 + 𝐴2)𝑦𝑡−1 − 𝐴2(𝑦𝑡−1 − 𝑦𝑡−2) + 𝐵𝑥𝑡 + 𝜃 + 𝜀𝑡 |−𝑦𝑡−1 (7)

𝑦𝑡 = (𝐴1 + 𝐴2 − 𝐼) 𝑦𝑡−1 − 𝐴2(𝑦𝑡−1 − 𝑦𝑡−2) + 𝐵𝑥𝑡 + 𝜃 + 𝜀𝑡, (8)

∆𝑦𝑡 = ∏𝑦𝑡−1 + 𝐴2∆𝑦𝑡−1 + 𝐵𝑧𝑡 + 𝜃 + 𝜀𝑡 (9)

where 𝐼 is the identity matrix.

For general p, VAR(p) can be written as

∆𝑦𝑡 = ∏𝑦𝑡−1 + ∑ Γ𝑖𝑝−1𝑖=1 ∆𝑦𝑡−𝑖 + 𝐵𝑧𝑡 + 𝜃 + 𝜀𝑡, (10)

where ∏ =∑ 𝐴𝑖𝑝𝑖=1 − 𝐼 (11)

and

Γ𝑖 = −∑ 𝐴𝑗𝑝𝑗=𝑖+1 . (12)

where, ∆ is the first-difference operator and Γ𝑖(𝑖 = 1, 2, … , 𝑝 − 1) are 𝑘 𝑥 𝑘

matrices of coefficients, where 𝑝 is the number of lags.

A result from linear algebra implies that, if the coefficient matrix ∏ has reduced

rank 0 < 𝑟 < 𝐾, then there exist 𝑘 𝑥 𝑟 matrices 𝛼 𝑎𝑛𝑑 𝛽 each with rank 𝑟 such

that ∏ = 𝛼𝛽′ and 𝛽′𝑦𝑡 is 𝐼(0). Granger’s representation theorem then states

that 𝛽 can be interpreted as co-integrating vectors.

Thus, equation (10) can be rewritten as:

∆𝑦𝑡 = 𝛼𝛽′𝑦𝑡−1 + ∑ Γ𝑖𝑝−1𝑖=1 ∆𝑦𝑡−𝑖 + 𝐵𝑧𝑡 + 𝜃 + 𝜀𝑡, (13)

where 𝛼𝛽′ = − Γ(1) = −𝐼𝑘 + Γ1 + Γ2+. . . +Γ𝑝

Furthermore, 𝑟 is the number of cointegrating vectors (the cointegration rank)

and each column of 𝛽 is the cointegration vector. One motivation for the VECM

(p) form is to consider the relation 𝛽′𝑦𝑡 = 𝑐 as defining the underlining economic

48

relations and assume that the agents react to the disequilibrium error 𝛽′𝑦𝑡 − 𝑐

through the adjustment coefficient 𝛼 to restore equilibrium; that is, they satisfy

the economic relations. The co-integrating vector 𝛽 is sometimes called the

long-run parameters. Thus, Johansen’s method is to estimate the ∏ matrix

from an unrestricted VAR and to test whether we can reject the restrictions

implied by the reduced rank∏.

Interpretation of ∆𝒚𝒕 = ∏𝒚𝒕−𝟏 + ∑ Γ𝒊𝒑−𝟏𝒊=𝟏 ∆𝒚𝒕−𝐢 + 𝑩𝒛𝒕 + 𝜺𝒕:

1. If ∏ = 0 𝑟𝑎𝑛𝑘 (∏) = 0, then there is no cointegration. Non-

stationarity of 𝐼(1) type vanishes by taking differences. That is 𝑦𝑡 are

stationary in difference.

2. If ∏ has full rank 𝑟𝑎𝑛𝑘 (∏) = k, then the 𝑦’𝑠 cannot be 𝐼(1) but are

stationary, so that all components of 𝑦𝑡 are 𝐼(0).

3. The most interesting case is 𝑟𝑎𝑛𝑘 (∏) = r , where 0 < 𝑟 < 𝑘, as this is

the case of cointegration. That is there are 𝑘 − 𝑟 linear combinations that

are non-stationary and 𝑟 stationary cointegrating relations. In other

words, we write

∏ = 𝛼𝛽′.

(𝑘𝑥𝑘) = (𝑘𝑥𝑟)[(kxr)′]

where the columns of 𝛽 contain the 𝑟 cointegrating vectors, and the

columns of 𝛼 the 𝑟 adjustment vectors.

Test for the Co-integration:

In order to distinguish between trend stationary and drift stationary series, the

Johansen-test (1995) is applied as follows:

1. The time series have linear trend and the co-integrating equations have

intercepts:

𝐻1∗(𝑟): ∏yt−1 + 𝐵𝑥𝑡 = 𝛼(𝛽′𝑦𝑡−1 + 𝑝0) + 𝛼 𝛾0

2. The time series have linear trend and the co-integrating equations have

linear trends:

𝐻∗(𝑟): ∏yt−1 + 𝐵𝑥𝑡 = 𝛼(𝛽′𝑦𝑡−1 + 𝑝0 + 𝑝1𝑡) + 𝛼 𝛾0

49

In order to test the number of co-integration vectors, the trace test (λ𝑡𝑟𝑎𝑐𝑒) and

the maximum-eigenvalue statistic (λ𝑚𝑎𝑥) are used in this model. The trace test

tests the null hypothesis of 𝑟 co-integrating vectors 𝑟 against the alternative

hypothesis of 𝑘 co-integration vectors, where 𝑘 is the number of endogenous

variables, for = 0, 1, … 𝑘 − 1, [𝐻0(𝑟) 𝑎𝑔𝑎𝑖𝑛𝑠𝑡 𝐻𝐴(𝑘)]. The alternative of 𝑘

cointegration relations corresponds to the case where none of the series has a

unit root. The maximum-eigenvalue test tests the null hypothesis of 𝑟 co-

integration vectors 𝑟 against the alternative hypothesis that there are 𝑟 + 1

cointegrating relationships between the series, [𝐻0(𝑟) 𝑎𝑔𝑎𝑖𝑛𝑠𝑡 𝐻𝐴(𝑟+1)]:

λ𝑡𝑟𝑎𝑐𝑒 (𝑟) = −𝑇∑ ln (1 − λ𝑖)𝑘𝑖=𝑟+1 (14)

λ𝑚𝑎𝑥 (𝑟, 𝑟 + 1) = −𝑇∑ ln (1 − λ𝑟+1)𝑘𝑖=𝑟+1 (15)

Here 𝑇 is the sample size and λ𝑖 is the i:th largest canonical correlation.

Note that in some cases trace and maximum-eigenvalue tests may yield

different results. Which of the two tests should be given priority, this is somehow

unsettled in the literature. Most authors tend to prefer the trace version.

50

9. CV

10

Cihan Yaylali, Bakk. rer. soc. oec Nationality: Turkish Marital status: single E-Mail: [email protected]

Ausbildung -Master in economics at the University of Vienna 2012-214 -Bachelor in economics at the Karl-Franzens-University in Graz 2007-2009 -Bachelor in constructional engineering at TU-Graz 2006-2007 -AHS-Matura BORG Dreierschützengasse Graz 2005

Publications

-In the modulus Gesundheit-und Pharmaökonomie: Bachelor thesis: "Patente in der Pharmabranche, 2012". -In the modulus Wachstum und Verteilung: Bachelor thesis: "Industriliazation and the Big Push, 2009".

„the relationship between immigration and unemployment...

Documents